Migrate from Antigravity to Xantly + OpenCode
Google cut Antigravity's free tier by 92% in December 2025. Replace it with OpenCode + Xantly in under 5 minutes and regain access to 10,000+ models.
On December 7, 2025, Google cut Antigravity's free tier from 250 requests/day to 20. Then on March 11, 2026 they introduced a credits system ($25 = 2,500 credits) and changed Pro's quota from 5-hour rolling windows to weekly, effectively another rate hike. The backlash is real; users are hunting for alternatives.
Here's how to replace Antigravity with OpenCode + Xantly in under 5 minutes. No Google account needed. Free tier stays generous. Your tokens, your margins.
What you're giving up (nothing worth keeping)
| Antigravity | OpenCode + Xantly |
|---|---|
| Gemini-only | 10,000+ models (Anthropic, OpenAI, Groq, NVIDIA, DeepSeek, Gemini, etc.) |
| Proprietary socket-TLS wire protocol | Standard HTTPS + OpenAI-compatible |
| 20 req/day free | Pay-as-you-go (typically $3-6/month for similar usage) |
| No BYOK | Full BYOK support, bring your OpenAI/Anthropic credits |
| Closed-source IDE | OSS CLI + full OSS stack |
| Locked to Google account | Any account |
What you get
- More models. BaRP routing across tiers, speed/value/quality auto-picked per request.
- Semantic cache. Repeated prompts (you re-ask variants all day) return near-instantly at near-zero cost.
- Memory across sessions. L3 graph memory remembers prior context.
- Waterfall fallback. One provider having a bad hour doesn't stall your flow.
- Real cost visibility. Every call logged with model, cache status, USD cost.
The 5-minute migration
Step 1, Install OpenCode (30 seconds)
npm install -g @opencode-ai/cli
Or use npx @opencode-ai/cli without installing.
Step 2, Get a Xantly API key (1 minute)
Sign up at xantly.com → Dashboard → API Keys → Create. Copy the sk-... value.
Step 3, Configure OpenCode (2 minutes)
Create ~/.config/opencode/opencode.jsonc:
{
"$schema": "https://opencode.ai/config.json",
"provider": {
"xantly": {
"npm": "@ai-sdk/openai-compatible",
"options": {
"baseURL": "https://api.xantly.com/v1",
"apiKey": "sk-..."
},
"models": {
"xantly/auto-quality": { "name": "Auto Quality (best T1)" },
"xantly/auto-value": { "name": "Auto Value (T2 balanced)" },
"xantly/auto-speed": { "name": "Auto Speed (lowest latency)" },
"bedrock/claude-sonnet-4-6/us-east-1": { "name": "Claude Sonnet 4.6" },
"bedrock/gpt-oss-120b-1/us-east-1": { "name": "GPT-5.4" },
"google/gemini-2.5-flash": { "name": "Gemini 2.5 Flash (if you still want Gemini)" }
}
}
},
"model": "xantly/auto-quality"
}
Step 4, Run (30 seconds)
cd your-project
opencode
Type a prompt. OpenCode routes through Xantly, which picks the best model, returns the answer, and logs the cost. Done.
Step 5, Set a budget cap (1 minute, optional but smart)
Xantly dashboard → API Keys → [your key] → Daily Budget USD: 5. Protects you from runaway costs while you adjust to the new workflow.
Cost comparison
Typical Antigravity Pro user (~100 req/day):
- Antigravity Pro: $20/month flat
- OpenCode + Xantly: $3-6/month depending on model mix (semantic cache typically cuts 30-50% off that)
Typical Antigravity Free user (capped at 20 req/day):
- Antigravity Free: $0 but 20 req/day ceiling
- OpenCode + Xantly with new-user credits: effectively $0 for the first ~300 requests, then ~$0.05/request after
You're not just escaping the tier cut, you're usually cheaper even vs Pro.
FAQ
Can I still use Gemini? Yes. Set your model to vertex/gemini-3.7-flash/global or google/gemini-2.5-flash and Xantly routes to Gemini directly. Or leave it on xantly/auto-quality and let BaRP pick the best option (which often includes Claude or GPT for quality reasons).
Do I need to rewrite any code? No. OpenCode is a separate IDE/CLI. If you have scripts hitting Antigravity's API directly, point them at https://api.xantly.com/v1 with your Xantly key, same OpenAI shape, drop-in.
What about the Antigravity-specific features (socket-TLS, Gemini Live)? Xantly doesn't speak the Antigravity wire protocol. If you need specific Gemini Live features, use the Gemini Live integration through Xantly's voice pipeline, it supports Gemini realtime directly.
My Google account has state in Antigravity. Export your conversation history from Antigravity (Settings → Export), then use the OpenCode @file mention to pull context into your first few sessions. After a week, memory will have absorbed the useful patterns.
Is OpenCode as good as Antigravity? Different product, different strengths. OpenCode is CLI-first and more developer-flexible. If you want a GUI IDE, try Cursor + Xantly, Zed + Xantly, or JetBrains AI Assistant + Xantly.
Next steps
- Use Xantly with OpenCode, deeper OpenCode config walkthrough.
- Cost-Optimized Routing, how BaRP keeps your bill down.
- Memory & Context, persistent project knowledge across sessions.
- Bring Your Own Key, if you have Gemini API credits, use them through Xantly.